A surprisingly large amount of Office for Student data activity is devoted to demonstrating how student characteristics (things generally beyond the control of the student) have an impact on student outcomes.

In some cases, the emphasis is on the need for providers to support students from all backgrounds in achieving good outcomes. It’s become normalised now but to imply that differences in outcomes based on characteristics was somehow the fault of providers was a decisive and controversial shift.

Previously, official statistics were used with benchmarks, to make allowances for the fact that students with different backgrounds had historically seen divergent outcomes. Outside of TEF, and September’s release of student characteristics data, such benchmarks are currently deemed to be excusing poor institutional performance.

The student characteristics data is awkward in this context as it eloquently makes the argument that student outcomes have a lot to do with characteristics and less to do with provider activity. And I’ve two rather complicated charts to help examine this.

What I’m trying to show in the first is the impact that characteristics have on the likelihood of a given student reaching any of the four OfS lifecycle milestones (continuation, completion, attainment, progression). The large number of blobs does make this look messy but it does help us to make some unusual comparisons (students studying in subcontractual arrangements tend to see worse progression than the most severely disadvantaged students) , but you can filter down to just one characteristic group using the blue boxes at the bottom. Do note that in some cases we get individual quintiles and groups of quintiles in the same chart – bad data design!

At the top, in the orange boxes, you can choose your lifecycle milestone of interest, plus the mode and level of study and a very broad (UK or non-UK) domicile – though note with the latter not all characteristics are available for non-UK students.

[Full screen]

This second chart works in the same way as the one above, but the “mode” selector is replaced by a choice of academic subjects. This helps us control for the impact of the subject of study, allowing us to see that in pretty much all subjects you have a better chance of progression to skilled employment if your family is rich.

[Full screen]

All this data, as you’ll no doubt have noted, is provided at a sector rather than an institution level. It is possible, to pick an example at random, that older students and students who live at their home address while studying (and is it possible that these are the same students?) tend to choose what we used to call “poor quality courses” at struggling providers.

John Blake’s Equality of Opportunity Risk Register (EORR) is equally clear that these are sector wide problems, and a recent funding circular appears to recognise that collaboration between and around providers is central to addressing systemic problems. No one provider, no one subject, is going to spontaneously beat systemic inequalities without a hell of a lot of support. Maybe there is a cost saving case to think about this nationally.